Diagnostics and prognostics utilising dynamic Bayesian networks applied to a wind turbine gearbox

Charles Edward Plumley, Graeme Wilson, Andrew Kenyon, Francis Quail, Athena Zitrou

Research output: Contribution to conferencePaper

3 Citations (Scopus)
379 Downloads (Pure)


The UK has the largest installed capacity of offshore wind and this is set to increase significantly in future years. The difficulty in conducting maintenance offshore leads to increased operation and maintenance costs compared to onshore but with better condition monitoring and preventative maintenance strategies these costs could be reduced. In this paper an on-line condition monitoring system is created that is capable of diagnosing machine component conditions based on an array of sensor readings. It then informs the operator of actions required. This simplifies the role of the operator and the actions required can be optimised within the program to minimise costs. The program has been applied to a gearbox oil testbed to demonstrate its operational suitability. In addition a method for determining the most cost effective maintenance strategy is examined. This method uses a Dynamic Bayesian Network to simulate the degradation of wind turbine components, effectively acting as a prognostics tool, and calculates the cost of various preventative maintenance strategies compared to purely corrective maintenance actions. These methods are shown to reduce the cost of operating wind turbines in the offshore environment.
Original languageEnglish
Number of pages11
Publication statusPublished - 12 Jun 2012
EventInternational Conference on Condition Monitoringand Machine Failure Prevention Technologies, CM & MFPT 2012 - London, United Kingdom
Duration: 12 Jun 201214 Jun 2012


ConferenceInternational Conference on Condition Monitoringand Machine Failure Prevention Technologies, CM & MFPT 2012
Country/TerritoryUnited Kingdom


  • diagnostics
  • prognostics
  • Bayesian networks
  • wind turbine

Cite this